Re: tmpfs race conditions

[Thomas Klausner <wiz%NetBSD.org@localhost>]

On Sat, Nov 28, 2015 at 09:38:57PM +0100, Joerg Sonnenberger wrote:
> On Mon, Oct 12, 2015 at 09:40:09PM +0200, Joerg Sonnenberger wrote:
> > I have a build machine building pkgsrc where I see some very strange
> > issues. The individual chroots share a read-only null-mounted base
> > directory with a number of tmpfs instances on top for most of the
> > writeable parts. Symptoms of the some data corruption or race
> > conditions:
> 
> I've seen one more strange situation. While trying to interactively
> debug a problem in an otherwise idle chroot, I've hit issues that look
> like the current directory disappeared under the process. At least some
> of the problems are visible on the releng build cluster as well, e.g.
> files missing during the check-flist stage that clearly have been
> installed before, impossible race conditions etc.

This reminds me of my PR 42484 from 2009, that I still see regularly today.
 Thomas